Neural tissue can be artificially recorded from and stimulated by prosthetic devices that sense or pass pulses of electrical current through electrodes on such a device. The passage of current causes changes in electrical potentials across neuronal membranes, which can initiate neuron action potentials, which are the means of information transfer in the nervous system.
Based on this mechanism, it is possible to read information from and input information into the central nervous system by coding or decoding the sensory information as a sequence of electrical pulses which are relayed to the central nervous system via the prosthetic device. In this way, it is possible to provide active motor prostheses and create artificial sensations.
In 1986, Bullara (U.S. Pat. No. 4,573,481) patented an electrode assembly for surgical implantation on a nerve. The matrix was silicone with embedded iridium electrodes. The assembly fit around a nerve to stimulate it.
US Patent Application 2003/0109903 to Berrang describes a Low profile subcutaneous enclosure, in particular and metal over ceramic hermetic package for implantation under the skin.
ECoG (electrocorticography) and LFP (local field potentials) have been shown to provide data useful for BMIs (Mehring 2004) with LFPs containing more information content, but with higher surgical risk. Also in 2004, an online study by Leuthardt et al. (Leuthardt 2004) showed that ECoG can support accurate BMI operation with little user training. Additionally, this study also provided initial evidence that ECoG signals contain information about the direction of hand movements in particular. This finding was important in revealing that high frequency gamma rhythms provide information not simply on focal cortical activations, but rather convey specific information about cognitive intent. Distinct from single unit studies, this is one of the earliest demonstrations that cognitive intent could be inferred from large population scale cortical physiology.
EEG is non-invasive and has supported important BMI applications, including two- and three-dimensional movement control (Farwell 1988 a, b; Wolpaw 1991 a, b, 1994, 2002, 2004; Sutter 1992; McFarland 1993, 2008, 2010; Pfurtscheller 1993; Birbaumer 1999; Kübler 1999, 2005; Pfurtscheller 2000; Millán 2004; Muller 2006; Vaughan 2006 Royer 2010). The highest functioning EEG-based BMIs, however, require a substantial degree of user training and their performance is often not reliable. BMIs that are based on intracortical recordings of action potential firing rates or local field potentials are on the opposite end of the performance and clinical spectrum (Georgopoulos 1986; Serruya 2002; Taylor 2002; Shenoy 2003; Anderson 2004; Lebedev 2005; Hochberg 2006; Santhanam 2006; Donoghue 2007; Velliste 2008). Though they can achieve a high level of multidimensional of control, there still remains a significant and unresolved question regarding the long-term functional stability of intracortical electrodes, particularly for recording action potentials (Shain 2003; Donoghue 2004; Davids 2006). This lack of signal durability has important clinical implications, because signal loss would require frequent replacement of the implant which would be neurosurgically unacceptable. Despite encouraging evidence that current non-invasive and invasive BMI technologies can actually be useful to severely disabled individuals (Kübler 2005; Hochberg 2006; Sellers 2010), these shortcomings and uncertainties remain substantial barriers to widespread clinical adoption and implementation in humans.
Compared to EEG, ECoG has major advantages: higher spatial resolution (i.e., 1.25 mm (subdural recordings (Freeman 2000; Leuthardt 2009) and 1.4 mm (epidural recordings (Slutzky 2010) vs. several centimeters for EEG); higher amplitude (i.e., 50-100 μV maximum vs. 10-20 μV maximum for EEG); far less vulnerability to artifacts such as electromyographic (EMG) or electroocular (EOG) activity ((Freeman 2003) or (Ball 2009), respectively); and broader bandwidth (i.e., 0-500 Hz (Staba 2002) vs. 0-40 Hz for EEG). With respect to the larger bandwidth of ECoG compared to EEG, it is important to note that this advantage may be directly related to the larger amplitude of ECoG. Because ECoG generally follows a 1/frequency drop-off in signal power (Miller 2009), task-related brain signals may remain larger than the noise floor of the amplifier/digitizer, and thus be detectable, at higher frequencies than for EEG. Additionally, these higher gamma frequencies (60-500 Hz) have been shown to carry substantive information on cognitive motor and language intentions and provide vital information for cognitive control features that are poorly accessible with EEG. In addition to these advantages of signal and information quality, ECoG electrodes (which do not penetrate cortex) should provide greater long-term functional stability (Pilcher 1973; Loeb 1977; Bullara 1979; Yuen 1987; Margalit 2003) than intracortical electrodes, which induce complex histological responses that may impair neuronal recordings. A recent study by Chao (2010) showed that the signal-to-noise ratio of ECoG signals, and the cortical representations of arm and joint movements that can be identified with ECoG, are stable over several months (Schalk 2010).
A fully integrated implantable architecture that combines electrode array, signal amplification, and telemetry has many advantages in creating a practical neural interface for a BMI based prosthetic limb system. Several such systems have been designed (Wise 2005; Harrison 2007; Rouse 2011). Among them the 100-electrode wireless cortical neural recording system based on the Utah array (Harrison 2007) represents the most up-to-date state-of-the-art in terms of electrode count, system integration and telemetry. It contains an array of 100 amplifiers (60 dB gain), a 10 bit ADC, an inductive power link, 20 kpbs forward telemetry data, and an FSK back telemetry link with a data rate of 345 kbps. However, this system is built to record from a spike electrode array that has shown problems of long term encapsulation and signal degradation due in-part to a mechanical stiffness mismatch. Additionally, the implant does not have a hermetic package with proven long-term reliability, like the Argus II. The newest implantable system intended for chronic BMI is the 16 electrode system built from the Medtronic ActivaPC neural stimulator by adding a “brain activity sensing interface IC” (Rouse 2011). By using ActivaPC's system scheme, the main device is to be implanted in a place on the body away from the head and the sensing electrode array to be connected to the device through an extension connector and subcutaneous cable. The neural sensing interface includes sub-μV resolution and is intended for both LFP and ECoG based control. However, it is a prototype system that is built for proof of concept and not a clinical device. Furthermore, the use of the extension connector and long subcutaneous cable is still susceptible to infection and cannot avoid the relatively common lead breakage problems associated with ActivaPC device (Hamani 2006; Blomstedt 2005; Fernandez 2010). And having the amplifiers so distant from the recording sites increases the introduction of noise. Also, the use of an implant battery limits its long term capabilities. And the long-leads limit MRI to low levels—an important clinical consideration.
No single state-of-the-art system contains all the key features necessary for a practical chronically implantable and reliable CNS-interface system with a long life.